Aboriginal Electoral Participation in Canada
2. Data and Methods
Our study relies on post-election surveys conducted by Elections Canada in 2004, 2006, 2008 and 2011. The surveys used random samples of adults in the Canadian population, with an oversample of Aboriginals.3 The total number of respondents was 2,822, 3,013, 3,348 and 3,570, respectively. Of these, 660, 637, 520 and 528 were Aboriginal people. All four surveys saw a mix of on-reserve and off-reserve respondents; on average, First Nations people living on reserves accounted for 41% of Aboriginal interviewees. Aboriginal respondents are well distributed across age groups and gender. Likewise, First Nations, Métis and Inuit respondents from all parts of Canada are included in the sample. One limitation, however, is that the sample includes a comparatively smaller number of Métis and Inuit respondents. We thus limit our analysis to Aboriginal respondents generally, while noting that future research should explore these different groups in more detail.4
Each survey included common demographic questions and a battery of common questions on political attitudes and beliefs. Each survey also included useful questions on registration and voter turnout. We take full advantage of all of these questions, wherever possible. While the surveys do not always include exactly the same questions, there is a sufficient common core of questions to directly compare the determinants of turnout across time.5
Our analyses employ standard methods from the social sciences, namely bivariate comparison and then multiple regression. We begin by comparing rates of turnout across a single dividing variable. We then perform multiple regressions to consider the independent effects of a variable on some outcome, while controlling for the effects of other variables. Multiple regression is a statistical technique that examines the relationship between a dependent variable (for example, height) and a number of independent variables (for example, parents' height, diet, exercise and gender). Rather than comparing the relationship between height and all of its possible causes separately, multiple regression considers all of these causes at the same time and determines the independent effect of each.
The estimated effect of each factor is represented by a regression coefficient. Coefficients tell us how strongly an independent variable is related to the dependent variable. Coefficients are accompanied by a p-value that tells us how sure we can be that the relationship between the two variables is not due to chance. Therefore, the larger the regression coefficient, the more important its effect. The smaller the p-value, the more certain we can be that the relationship is real and not due to chance. We say that a relationship that is not due to chance is statistically significant.
3 The survey response rates were 28%, 32%, 26% and 25%, respectively.
4 Full sample details are available upon request.
5 Full surveys are likewise available upon request.